British Journal of Radiology (2003) 76, S92-S96
© 2003 British Institute of Radiology
doi: 10.1259/bjr/25400972
Imaging in antiangiogenesis trial: a clinical trials radiology perspective
M Kothari, PhD
1
A Guermazi, MD
1,2
D White, PhD
1
J Suhy, PhD
1 and
C Reinhold, MD
1,3
1 Synarc Inc., 575 Market St Suite 1700, San Francisco CA 94117, 2 University of California at San Francisco, 350 Parnassus Avenue, Suite 150, San Francisco, CA 94117, USA and 3 McGill University Health Center, 1650 Cedar Ave, Montreal, Quebec, Canada, H3G1A4
Correspondence: Dr Manish Kothari, VP, Medical Imaging Technology, 575 Market St Suite 1700, San Francisco CA 94105, USA
 |
Abstract
|
|---|
Traditional approaches for treating cancer have largely focused on the ability of chemotherapy, and to a lesser extent radiation therapy, to destroy tumour cells. Recent developments in antiangiogenesis treatments require a fundamental shift in the radiological and imaging paradigms associated with evaluating response. Proper design and execution of any clinical trial involving imaging angiogenesis requires satisfactory consideration of a number of strategies and an in-depth understanding of different imaging techniques such as dynamic contrast enhanced MRI and CT, contrast-enhanced ultrasound and positron emission tomography. In particular, for imaging, the strategies can be divided into issues that need to be addressed during the protocol planning phase, and strategies that need to be addressed during the execution phase. Furthermore, clinical trials are usually subject to stringent regulations surrounding traceability and reproducibility that need to be followed before the regulatory authorities will accept the integrity of the data. This paper elaborates on the above strategies and outlines certain aspects, or tactics, that need to be considered while preparing for a multicentre clinical trial that involves imaging angiogenesis.
|
Introduction
|
|---|
Traditional approaches for treating cancer have largely focused on the ability of cytotoxic agents, and to a lesser extent radiation therapy, to destroy tumour cells. These approaches have been tested in large multicentre clinical trials using morphological criteria such as World Health Organization (WHO) and the Response Evaluation Criteria in Solid Tumours (RECIST) to measure tumour response and progression [1]. Recent developments in antiangiogenesis treatments represent a shift in paradigms to treat cancer [2] – they also require a fundamental shift in the radiological and imaging paradigms associated with evaluating response [3].
Clinical trials radiology often represents the first extension of these new imaging techniques and occupies a niche between clinical practice and academic research. Academic research focuses on establishing the scientific validity of cutting-edge techniques. Often "enhanced" imaging hardware (and software) with abilities superior to standard clinical imaging hardware (and software) is used, and often longer imaging times are required. Clinical practice, on the other hand, requires that the imaging technique be widely available, reproducible at different sites, stable over time, easy to perform, low in cost, and provide maximum patient comfort and compliance [4]. Clinical trials radiology borrows traits from both academic radiology and clinical practice; however, its drivers and objectives make it a unique class by itself. In clinical trials, the patient is an abstraction, and the objective is to demonstrate formally the efficacy, safety and cost utility of a new therapy for regulatory approval, with the fewest patients and in the shortest time possible. Interestingly enough, it is precisely the fact that clinical trials have shared traits with academic radiology and clinical practice that often lead to misconceptions that hamper the design and execution of the trial, from an overestimation of the precision of a marker (and thereby an underestimation of the power of the technique), to allowing the sites performing complicated imaging procedures to be dictated solely by insurance provider considerations.
Proper design and execution of any clinical trial involving imaging requires satisfactory consideration of a number of strategies. In particular, for imaging, the strategies can be divided into issues that need to be addressed during the protocol planning phase, and issues that need to be addressed during the execution phase. The following sections elaborate on the above strategies and outline certain aspects, or tactics, that need to be considered while preparing for a clinical trial that involves imaging angiogenesis.
|
Protocol design phase
|
|---|
Understand the economic constraints
Large clinical trials in oncology often take advantage of the local healthcare and insurance systems so that the sponsor does not have to pay for costs that may also be associated with standard patient care. For oncology patients, this standard care often includes imaging. As a result, the choice of imaging sites to be used in the trial is typically not in the hands of the principal investigator. Furthermore, patients enrolled in the trial may end up being imaged at a number of different sites depending on insurance provider preferences and proximity to the patient's home. Although the quality of images obtained in this setting may be sufficient for relatively simple endpoint measurements such as lesion size, they are clearly inadequate when using more complex techniques used in antiangiogenesis trials such as dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), dynamic contrast-enhanced computed tomography (DCE-CT) or positron emission tomography (PET) imaging.
Select the correct imaging markers to assess the therapy
Although the primary rationale behind the choice of the imaging markers should be disease- and drug-mechanism, it is important to factor in logistical and recruitment practicalities in this decision. For example, some regions of the body are less amenable to DCE-MRI than others. While such structures as the bone, breast, and brain are considered stationary and thus relatively easy to scan, lesions in the lung and liver are more challenging because of respiratory motion, which can result in poor registration of a time sequence of images [3]. This situation is further aggravated in "real world" patients where respiratory problems due to the clinical status or the location and/or size of the tumour are not uncommon. If liver and lung metastases are to be imaged using DCE-MRI, a moving region-of-interest analysis might be required rather than a pixel-by-pixel evaluation of tumour contrast uptake [5]. This will impact the analysis plan: will moving whole tumour region-of-interest analysis be performed on all tumours, or only on lung and liver, with the remainder having the more sophisticated pixel-by-pixel analysis? What impact will these choices have on sample size, patient recruitment and drug labelling? Are there side-effects of the contrast that need to be considered? Should there be exclusion criteria in place to filter out patients at high risk for experiencing those side effects?
Therefore, in the context of a multicentre clinical trial, it is important to consider all possible modalities and imaging markers, since DCE-MRI, PET, DCE-CT and ultrasound all have their strengths and limitations [6, 7]. Care should be taken that the scientific advisors represent this range of imaging markers. Furthermore, their implementation of a technique may also not be standard — it is important to use published consensus protocols or commonly accepted protocols. Lung tumours that cannot be imaged adequately by DCE-MRI because of motion artefact may be able to be imaged by DCE-CT. If both CT and PET are to be used, MRI markers such as standardized perfusion value (SPV – a parameter that corrects perfusion values for patient weight and cardiac output) could be used for the MR analyses. The SPV has a similar mathematical derivation to the standardized uptake value (SUV) used to quantify tumour uptake of fluorodeoxyglucose during PET. It may be possible to develop a statistical plan that includes analyses from multiple modalities if early consideration is given to the panel of imaging markers [8].
Understand the precision and accuracy of the marker in a multicentre setting
The reproducibility, or precision, of an imaging marker is often required to estimate the sample size required to evaluate therapy efficacy. In the case of newer techniques, the precision estimates are quite often obtained from single centre studies at premier research institutions. This may result in the trial being insufficiently powered, since the precision of an imaging marker is usually better in a carefully controlled setting than that typically seen in a multicentre clinical trial. This is a difficult situation, since accurate multicentre precision numbers are usually unavailable, or at best represent multicentre precision seen in small studies performed at premier institutions. This situation is particularly acute when techniques are beginning their migration from academic radiology to clinical trials. This is the case for most angiogenesis imaging approaches. The solution lies in invoking an early dialogue between the academic thought leaders and the scientists and radiologists in the clinical trials group. Preferably this dialogue should take place early, during the protocol design phase. The clinical trials radiology group may have access to precision data from other, similar trials. They may be able to suggest self-calibration approaches that have been accepted by regulatory authorities (such as per patient test–retest precision) as a way of assessing significant change. They may be able to suggest randomized trial designs that allow for ongoing evaluation of precision, yet can be powered up to ensure adequate sample size once precision information is available. Finally, for pivotal studies, they may be in a position to rapidly carry out a multicentre study to evaluate the reproducibility.
|
Study planning and execution phase
|
|---|
Imaging facility selection
From an imaging perspective, selection of the imaging facility is a critical step in optimizing accuracy and minimizing variance. However, from a sponsor perspective, patient recruitment is the key step, and the choice of the clinical site and principal investigators is also critical. The imaging facility is often a secondary consideration. This dilemma often hampers the ability to successfully incorporate advanced imaging techniques into clinical trials.
The degree to which imaging equipment across the multiple sites included in a clinical trial need to be standardized depends on the type of measurements that will be made, and the ability of the central radiology service to deal with multivendor image data. Ideally, all equipment, software platforms and upgrade schedules at all of the sites throughout the duration of the study should be identical, but this is rarely feasible. The choice of the imaging site is usually determined by proximity to the recruiting centre, and patient comfort and convenience. However, early involvement of imaging specialists can help qualify the hardware and software and monitor upgrades throughout the study. It is important to qualify imaging sites early during the planning phase of the trial. Sometimes, the clinical trials radiology groups may be able to suggest sites that they have worked with. It is better to not include an inappropriately accoutred site than to raise the variance and reduce the accuracy of the study by including the site.
The factors involved in this are:
- Understanding the limitations of different imaging system makes (e.g. GE, Siemens, Philips, Toshiba, etc.) and models (GE5.x versus GE LX, for example)
- Build acceptable flexibility in the imaging protocol to allow for differing imaging equipment from a variety of manufacturers.
- Perform a phantom test run to determine instrument and software capabilities and adherence to protocol prior to patient scans.
- Understanding the qualification and motivation of technologists who will perform individual studies.
It is important to note that a Siemens MR scanner is not a General Electric (GE) MR scanner is not a Philips MR scanner. Even the same manufacturer will have important differences between their various platforms even in the same model. These relate to both hardware and software and will affect the performance for the simplest examinations. These differences are not just cosmetic but affect the type of image acquisitions that can be performed and therefore the ultimate analysis plan. For example, navigator echoes are available in newer Siemens and Philips MR scanners. These are not yet available on standard GE scanners or on older Siemens and Philips scanners. Navigator echoes can correct for small motion artefacts and make it possible to perform pixel-by-pixel analyses using DCE-MRI on certain liver and lung metastases [3]. However, if some of the sites have navigator echoes and others do not, they may be required not to use them in order to ensure a consistent analysis through all sites. The exact matrix size may never be the same between different makes. The section thickness may vary for a given sequence. Siemens multidetector CT uses adaptive array technology, whereas GE Light speed 16 uses fixed array technology. Both have their advantages, but only serve to increase the variance of the analysis. PET scanners from different manufacturers have different constraints, and some will be more precise or have more features than others for certain markers. Knowledge of the site capabilities early on in the study can help define the patient inclusion/exclusion criteria, and thereby the protocol and analysis plan. Another example relates to the availability of power injectors. Power injectors allow precise application of contrast bolus volume and bolus rate of delivery, followed by a precise volume of saline wash. It is well known that the substantial variations in the precision of a calculated kinetic parameter can occur when using a manual injection technique which is related to the difficulty in obtaining a reproducible and uniform rate of contrast flow; this can be minimized by a practiced contrast material administrator who follows a written, standardized injection protocol. Because a slower or faster injection rate leads to a decreased qualitative and quantitative image quality, the use of an automatic MR power injector is superior to manual injection of contrast material. For these reasons, multicentre studies not employing power injectors invariably result in lower longitudinal reproducibility.
Although appropriate imaging hardware is a crucial minimum requirement, without a doubt, the easiest and most effective way to control longitudinal variance is to use experienced technologists. The best technologists produce the best images. It is important to establish their experience and credentials. Providing specific training of these technologists on the intricacies of the specific imaging protocol improves the quality of the images. It is also necessary to take the effort to explain the value and utility of quality assurance (QA) protocols. Efforts to include the technologist as a pivotal member of the clinical trial team will improve motivation and maximize compliance with the imaging protocol. Although many of the techniques used to image angiogenesis can be performed relatively easily using existing imaging hardware, the procedures differ considerably from most conventional techniques. Greater care must be exercised in pre-contrast scanning, injection rate and dose, image timing and image analysis. Some of the newer machines allow the technologist to pull up previous dynamic images of the patient and apply the exact protocol to the current time point. Utilizing these new features can dramatically increase the consistency of the imaging protocol over time. Working with a qualified and motivated technologist can enable use of these features.
Planning for phantom usage
A small but critical aspect of any imaging methodology is the use of phantoms in the clinical trial. Phantoms are used for a variety of reasons – to monitor machine drift, to calibrate the images, and to correct variations in section thickness, contrast or other features from time point to time point. Whatever the reason, the following rules apply:
- Phantoms cost money: plan for the phantom costs in your budget. A good T1 calibration phantom can cost several thousands of dollars each – if you need to buy one for each site, the cost can add up. If phantoms are going to be shipped from site-to-site, then shipping and insurance costs should be factored in.
- Phantoms take time to build and deliver: do not expect the phantom to arrive at the site immediately – they sometimes take up to 3 months from date of order to arrive. If the delivery is international, there may be further delays because of customs.
- Phantoms take imaging time: this is a crucial point that is often overlooked. If a phantom scan is being performed with each patient visit, it may add 10–15 minutes of scan time. This needs to be budgeted with the site during the scan cost negotiation process. It is not uncommon for this to become a point of contention later on in the trial.
- The phantom protocol must be simple, comprehensive but purposeful.
- The failure of a phantom examination to meet quality criteria should immediately prompt a number of questions, e.g. Which patient exams are affected? Can they be corrected, or must they be discarded? Does the system require servicing before more patients are imaged? A decision tree addressing all such anticipated factors should be established before the study begins.
Quality assurance
Once the imaging sites have been selected, the imaging protocol designed, and the technologists trained, image acquisition, transfer and quality must be closely supervised to ensure a high-quality image set for analysis. Variability in image quality can be introduced either by the manner in which the subject is prepared for the examination and/or by improper calibration and maintenance of the imaging system. Scanner performance is documented and maintained by performing scanner quality control. One aspect of this is done by the imaging sites as part of their routine clinical quality control, but additional study-specific quality control must also be performed for clinical trials. As mentioned above, this often requires the use of specialized phantoms.
After the images are acquired according to the study-specific protocol, they are transmitted to the central radiologist(s) for review of protocol compliance, patient positioning, anatomical coverage, and image quality. This requires explicit image quality (IQ) criteria. Computer programs can automatically check parameters that explicitly appear in the image file headers. However, other criteria such as patient positioning, anatomical coverage, absence of artefacts, and overall image quality have to be evaluated by a radiologist or other trained individuals. Protocols involving contrast agents usually involve additional layers of complexity, such as subtraction of pre- and post-contrast images to detect patient motion (Figure 1
), or examination of signal intensity versus time curves to evaluate the quality of the contrast agent bolus. If the images are of acceptable quality, they are entered into the central study database. All processes performed by the central radiology service must be done in strict accordance with standard operating procedures (SOPs) and study specific procedures (SSPs). If the image data do not pass the incoming quality inspection, a decision needs to be made as to whether or not the data can be corrected, for example, by using information obtained from the instrument quality control. If the images cannot be salvaged, the imaging must be repeated or the data point discarded. The criteria applied are again controlled by SOPs, and data exclusions must be carefully documented. If repeat imaging is necessary a feed-back loop must be designed with appropriate timing criteria. The ultimate result of this process is a high-quality image dataset that conforms to rigorous QA principles and can support high-quality and complex image analysis.
View larger version (65K):
[in this window]
[in a new window]
|
Figure 1. MRI of a brain tumour. The left and centre panels show pre- and 12 min post-injection T1 weighted spin-echo images. The right panel is the difference image. The slight mis-registration is most evident in the lateral anterior aspect and the margins of the sulci.
|
|
Centralized image analysis
In contrast to image acquisition, which must be performed at multiple sites, image analysis for clinical trials can be centralized. This allows the use of specially trained readers and dedicated software to achieve superior results to those that would be attainable with decentralized analysis. There is increased expectancy by regulatory agencies that the readers be independent (not participating in the study, not affiliated with the sponsor, and not affiliated with the institutions at which the study was conducted) and blinded (unaware of treatment identity, unaware – or have limited awareness – of patient-specific clinical information or of the study protocol, and not familiar with the inclusion and exclusion criteria for patient selection specified in the protocol). If necessary, large reading loads can be divided among multiple readers to accelerate the rate of analysis, but it is important that the same reader analyse all images in an individual patient series. Additionally, each reader should analyse the images independently of the other blinded readers and of any on-site readings performed by the clinical investigators or radiologists at the imaging sites [4].
Consistency among the readers should be quantified using statistics appropriate for the nature of the data. Typically, percentage agreement or kappa statistic are used for nominal/existential markers; percentage agreement, weighted kappa statistic or non-parametric correlation indices (Spearman's rho, Kendal's tau) are used for ordinal, or ranked categorical markers; and Pearson correlation coefficient, coefficient of variation, or intraclass correlation coefficient are used for dimensional or continuous markers.
Imaging markers for angiogenesis are novel and still emergent. Accordingly, it is quite possible that the techniques being suggested for use have achieved scientific validity in single-centre settings, but are not proven in multicentre settings. Consideration must be paid to the fact that software may have to be created and validated to meet the needs of the multicentre studies.
Regulatory compliance
The data collection and analysis will need to meet all aspects of regulatory compliance, as well as all applicable International Council on Harmonization (ICH) and Good Clinical Practice (GCP) rules. All personnel associated with the trials may need to have documented training in ICH and GCP. Any software used for the study would need to meet a variety of regulatory requirements such as the FDA's 21 CFR Part 11 on electronic signatures and audits in the database [www.fda.gov/cdrh/comp/guidance/]. These requirements establish criteria for software validity, data traceability and reproducibility. Guidances issued by the EMEA and FDA also establish requirements for the storage of the data, and standards for the number of years that the analysis centre needs to store electronic and paper records pertaining to image collection and data analysis. All these restrictions place a burden on the execution of a clinical trial, and require advance preparation and planning.
|
Conclusions
|
|---|
Clinical trials in angiogenesis imaging present new but familiar challenges to the field of clinical trials radiology. There is rapid emergence of innovative image acquisition techniques and of image analysis methodologies for monitoring disease progression and response to therapy using antiangiogenic agents. There is a desire to quickly move a technique from academic centres to a multicentre trial. The key to success in these processes includes the engagement of clinical trials radiology group with appropriate experience from trial design, QA, with proven data trace ability capability, and the ability to deliver at a level acceptable to regulatory authorities. An active dialogue of all concerned parties (sponsors, clinicians, images, technologists, regulators) is key to the successful implementation of the angiogenesis imaging techniques.
|
References
|
|---|
- Therasse P, Arbuck SG, Eisenhauer EA, Wanders J, Kaplan RS, Rubinstein L, et al. New guidelines to evaluate the response to treatment in solid tumors. European Organization for Research and Treatment of Cancer, National Cancer Institute of the United States, National Cancer Institute of Canada. J Natl Cancer Inst 2000;92:205–16.[Abstract/Free Full Text]
- Rosen L. Antiangiogenic strategies and agents in clinical trials. Oncologist 2000;5:20–7.[Abstract/Free Full Text]
- Padhani AR, Neeman. M Challenges for imaging angiogenesis. Br J Radiol 2001;74:886–90.[Free Full Text]
- Peterfy CG. Magnetic resonance imaging of rheumatoid arthritis: the evolution of clinical applications through clinical trials. Semin Arthritis Rheum 2001;30:375–96.[CrossRef][Medline]
- Choyke PL, Dwyer AJ, Knopp MV. Functional tumor imaging with dynamic contrast-enhanced magnetic resonance imaging. J Magn Reson Imaging 2003;17:509–20.[CrossRef][Medline]
- Bobek-Billewicz B, Szurowska E, Zapasnik A, Izycka-Swieszewska E, Gorycki T, Nowakowski M. Localisation of focal liver lesions to specific hepatic segments--comparison of multiphase spiral CT and MR imaging. Folia Morphol (Warsz) 2002;61:291–7.
- Yang M, Martin DR, Karabulut N, Frick MP. Comparison of MR and PET imaging for the evaluation of liver metastases. J Magn Reson Imaging 2003;17:343–9.[CrossRef][Medline]
- Miles KA. Functional computed tomography in oncology. Eur J Cancer 2002;38:2079–84.
Top
Abstract
Introduction
